Australian defence experts have been hard at work developing responsible artificial intelligence defence systems, galvanised by increasing geopolitical friction during Russia’s full-scale invasion of Ukraine. “The nature of modern conflict, with the overwhelming amount of information for the warfighter and the use of autonomous systems, brings to the fore the need for advanced technologies such as AI and autonomous systems,” says Dr Robert Hunjet, the Defence Science and Technology Group’s (DSTG) Program Lead of Artificial Intelligence.
“Defence is committed to using AI-enabled technology responsibly, with careful consideration of risks and potential impacts in line with Australia’s obligations under international humanitarian law.” AI has enormous defence potential. Last year, defence experts from the three AUKUS nations got together for an advanced capabilities AI and autonomy working group trial in the UK.
Dr Hunjet says they shared data, created AI models and uploaded them on one another’s platforms to generate the ability to quickly detect objects such as land vehicles and tanks in the given terrain.
The AI trial tested the object identification capabilities of AI models in a range of military vehicles, including Australian Insitu CT220 uncrewed aerial vehicles (UAVs or drones) and the UK’s Blue Bear Ghost UAVs. Using data collected from overhead images and photos of different types of land vehicles, Dr Hunjet’s team and international colleagues were able to train the AI models to rapidly detect the land vehicles. The experts managed to rapidly retrain models and uploaded these while the UAVs were in flight.
“We were able to really quickly turn around new models to detect objects that we weren’t able to detect before,” Dr Hunjet says. “It’s a really effective way of gaining situational awareness needed by the warfighters, at speed, so we have a better picture of the battlespace.”
These advanced capabilities enable Australia to maintain “collective multi-domain awareness, operate seamlessly with partners and enhance peace, security and prosperity in the Indo-Pacific”, Defence says.
In a separate activity in late 2023, the AUKUS partners joined forces for the TORVICE trial in South Australia. The TORVICE trial tested the resilience of autonomous land vehicles operating in an environment where they were subject to attacks from electronic warfare, electro optical laser, and position, navigation and timing systems.
“Any future conflict is going to be in a contested environment,” Dr Hunjet says. “GPS may be denied, communications may be denied.”
The trial was designed to find any gaps or vulnerabilities in the technology. This cooperative AUKUS work is primarily an accelerator, bringing experts from the US, the UK and Australia together to join forces and achieve better and faster results with artificial intelligence systems.
“We need to realise our competitors are progressing at pace and we need to do the same,” Dr Hunjet says.
As part of its commitment to responsible use of AI, Australia publicly endorsed the first-of-type Responsible AI in the Military Domain Summit Call to Action on February 16, 2023, and the Political Declaration on Responsible Military Use of Artificial Intelligence and Autonomy announced by the Deputy Prime Minister Richard Marles and Foreign Minister Penny Wong on November 3, 2023.
Dr Hunjet says defence research was driven by an appreciation of AI’s ability to sort through huge volumes of data at speed.
“We’re looking at using AI to enable efficiencies in identifying objects of interest, to process large volumes of information at pace, and to gain an asymmetric advantage through the use of these technologies,” he says.
AI can be used to coordinate “swarming”, which Dr Hunjet said is also known as multi-agent teaming, or machine-to-machine teaming, where many machine operatives self-coordinate within a military operation.
The AI-enabled systems can use auction-based algorithms for task allocation, in which the machines send messages to one another and bid on the tasks they can do.
“The swarm would not be deployed to operate completely autonomously; the operator is part of the human-machine team,” he says.
Experts from both the commercial and academic worlds are working with Defence through two key initiatives to build Australia’s capacity in AI in order to support the increasing number of defence projects that require AI, Dr Hunjet says.
The first initiative is the Centre for Advanced Defence Research in Robotics and Autonomous Systems (CADR-RAS), based at the University of Adelaide which draws on the expertise of researchers from 10 universities. The second initiative is the Australian Defence Artificial Intelligence research network (DAIRNet), based at the University of South Australia, brings together experts from defence, academia and industry as a focal point for academic engagement on Defence AI.